MIMIC models for uniform and non-uniform DIF as moderated mediation models

In this note we describe how multiple indicators multiple cause (MIMIC) models for studying uniform and non-uniform differential item functioning (DIF) can be conceptualized as mediation and moderated mediation models. Conceptualizing DIF within the context of a moderated mediation model helps us understand DIF as the effect of some variable on our measurements which is not accounted for by the latent variable of interest. In addition, this allows us to apply useful concepts and ideas from the mediation and moderation literature: (1) improving our understanding of uniform and non-uniform DIF as direct effects and interactions, (2) understanding the implication of indirect effects in DIF analysis, (3) clarifying the interpretation of the “uniform DIF parameter” in the presence of non-uniform DIF, and (4) probing interactions and using the concept of “conditional effects” to better understand the patterns of DIF across the range of the latent variable.

Te Waka Kuaka, Rasch analysis of a cultural assessment tool in traumatic brain injury in Māori

Background: The aim was to examine the validity of a new measure, Te Waka Kuaka, in assessing the cultural needs of Māori with traumatic brain injury (TBI). Methods: Māori from around Aotearoa, New Zealand were recruited. 319 people with a history of TBI, their whānau (extended family members), friends, work associates, and interested community members participated. All completed the 46-item measure. Rasch analysis of the data was undertaken. Results: All four subscales; Wā (time), Wāhi (place), Tangata (people) and Wairua practices (activities that strengthen spiritual connection) were unidimensional. Ten items were deleted because they did not fit the model, due to statistically significant disordered thresholds, non-uniform differential item functioning (DIF) and local dependence. Five items were re-scored in the fourth subscale resulting in ordered thresholds. Conclusions: Rasch analysis facilitated a robust validation process of Te Waka Kuaka.

Performance on the Boston Naming Test in Bilinguals

Objectives: We examined performance on the Boston Naming Test (BNT) in older and younger adults who were monolingual English or French speakers, or bilingual speakers of English and French (n=215). Methods: Monolingual participants completed the task in their native language, and bilingual participants completed the task in English, French, and bilingual (either-language) administrations. Results: Overall, younger and older monolingual French speakers performed worse than other groups; bilingual participants performed worst in the French administration and approximately two-thirds of bilingual participants performed better when responses were accepted in either language. Surprisingly, however, a subset of bilinguals performed worse when responses were accepted in either language as compared to their maximum score achieved in either English or French. This either-language disadvantage does not appear to be associated with the degree of balanced bilingualism, but instead appears to be related to overall naming abilities. Differential item analysis comparing language groups and the different administrations identified several items that displayed uniform and/or non-uniform differential item functioning (DIF). Conclusions: The BNT does not elicit equivalent performance in English and French, even when assessing naming performance in monolingual French speakers using the French version of the test. Scores were lower in French overall, and several items exhibited DIF. We recommend caution in interpreting performance on these items in bilingual speakers. Finally, not all bilinguals benefit from an either-language administration of the BNT. (JINS, 2015, 21, 350–363)

Establishing Effect Size Guidelines for Interpreting the Results of Differential Bundle Functioning Analyses Using SIBTEST

The purpose of this simulation study was to establish general effect size guidelines for interpreting the results of differential bundle functioning (DBF) analyses using simultaneous item bias test (SIBTEST). Three factors were manipulated: number of items in a bundle, test length, and magnitude of uniform differential item functioning (DIF) against the focal group in each item in a bundle. A secondary purpose was to validate the current effect size guidelines for interpreting the results of single-item DIF analyses using SIBTEST. The results of this study clearly demonstrate that ability estimation bias can only be attributed to DIF or DBF when a large number of items in a bundle are functioning differentially against focal examinees in a small way or a small number of items are functioning differentially against focal examinees in a large way. In either of these situations, the presence of DIF or DBF should be a cause for concern because it would lead one to erroneously believe that distinct groups differ in ability when in fact they do not.

Uniform DIF and DIF Defined by Differences in Item Response Functions

Uniform differential item functioning (DIF) exists when the statistical relationship between item response and group is constant for all levels of a matching variable. Two other types of DIF are defined based on differences in item response functions (IRFs) among the groups of examinees: unidirectional DIF (the IRFs do not cross) and parallel DIF (the IRFs are the same shape but shifted from one another by a constant, i.e., the IRFs differ only in location). It is shown that these three types of DIF are not equivalent and the relationships among them are examined in this paper for two item response categories, two groups, and an ideal continuous univariate matching variable. The results imply that unidirectional and parallel DIF which have been considered uniform DIF by several authors are not uniform DIF. For example, it is shown in this paper that parallel three-parameter logistic IRFs do not result in uniform DIF. It is suggested that the term “uniform DIF” be reserved for the condition in which the association between the item response and group is constant for all values of the matching variable, as distinguished from parallel and unidirectional DIF.